Datasets:
Tasks:
Summarization
Modalities:
Text
Formats:
csv
Languages:
English
Size:
10K - 100K
ArXiv:
License:
knkarthick
commited on
Commit
•
6a1510b
1
Parent(s):
9387aaa
Update README.md
Browse files
README.md
CHANGED
@@ -27,54 +27,48 @@ pretty_name: SAMSum Corpus
|
|
27 |
# Dataset Card for SAMSum Corpus
|
28 |
## Dataset Description
|
29 |
### Links
|
30 |
-
- **Homepage:**
|
31 |
-
- **Repository:** https://
|
32 |
-
- **Paper:** https://
|
33 |
- **Point of Contact:** https://huggingface.co/knkarthick
|
34 |
|
35 |
### Dataset Summary
|
36 |
-
|
|
|
|
|
37 |
### Languages
|
38 |
English
|
39 |
|
40 |
## Dataset Structure
|
41 |
### Data Instances
|
42 |
-
|
43 |
The first instance in the training set:
|
44 |
-
{'id': '
|
|
|
45 |
### Data Fields
|
46 |
- dialogue: text of dialogue.
|
47 |
-
- summary: human written summary of the dialogue.
|
48 |
-
- topic: human written topic/one liner of the dialogue.
|
49 |
- id: unique file id of an example.
|
50 |
|
51 |
### Data Splits
|
52 |
-
- train:
|
53 |
-
- val:
|
54 |
-
- test:
|
55 |
-
- holdout: 100 [Only 3 features: id, dialogue, topic]
|
56 |
|
57 |
## Dataset Creation
|
58 |
### Curation Rationale
|
59 |
In paper:
|
60 |
-
|
61 |
-
|
62 |
-
Compared with previous datasets, dialogues from DialogSum have distinct characteristics:
|
63 |
|
64 |
-
Under rich real-life scenarios, including more diverse task-oriented scenarios;
|
65 |
-
Have clear communication patterns and intents, which is valuable to serve as summarization sources;
|
66 |
-
Have a reasonable length, which comforts the purpose of automatic summarization.
|
67 |
-
|
68 |
-
We ask annotators to summarize each dialogue based on the following criteria:
|
69 |
-
Convey the most salient information;
|
70 |
-
Be brief;
|
71 |
-
Preserve important named entities within the conversation;
|
72 |
-
Be written from an observer perspective;
|
73 |
-
Be written in formal language.
|
74 |
### Who are the source language producers?
|
75 |
linguists
|
76 |
### Who are the annotators?
|
77 |
language experts
|
|
|
|
|
|
|
|
|
78 |
|
79 |
## Licensing Information
|
80 |
non-commercial licence: CC BY-NC-ND 4.0
|
|
|
27 |
# Dataset Card for SAMSum Corpus
|
28 |
## Dataset Description
|
29 |
### Links
|
30 |
+
- **Homepage:** hhttps://arxiv.org/abs/1911.12237v2
|
31 |
+
- **Repository:** https://arxiv.org/abs/1911.12237v2
|
32 |
+
- **Paper:** https://arxiv.org/abs/1911.12237v2
|
33 |
- **Point of Contact:** https://huggingface.co/knkarthick
|
34 |
|
35 |
### Dataset Summary
|
36 |
+
The SAMSum dataset contains about 16k messenger-like conversations with summaries. Conversations were created and written down by linguists fluent in English. Linguists were asked to create conversations similar to those they write on a daily basis, reflecting the proportion of topics of their real-life messenger conversations. The style and register are diversified - conversations could be informal, semi-formal or formal, they may contain slang words, emoticons and typos. Then, the conversations were annotated with summaries. It was assumed that summaries should be a concise brief of what people talked about in the conversation in third person.
|
37 |
+
The SAMSum dataset was prepared by Samsung R&D Institute Poland and is distributed for research purposes (non-commercial licence: CC BY-NC-ND 4.0).
|
38 |
+
|
39 |
### Languages
|
40 |
English
|
41 |
|
42 |
## Dataset Structure
|
43 |
### Data Instances
|
44 |
+
SAMSum dataset is made of 16369 conversations distributed uniformly into 4 groups based on the number of utterances in con- versations: 3-6, 7-12, 13-18 and 19-30. Each utterance contains the name of the speaker. Most conversations consist of dialogues between two interlocutors (about 75% of all conversations), the rest is between three or more people
|
45 |
The first instance in the training set:
|
46 |
+
{'id': '13818513', 'summary': 'Amanda baked cookies and will bring Jerry some tomorrow.', 'dialogue': "Amanda: I baked cookies. Do you want some?\r\nJerry: Sure!\r\nAmanda: I'll bring you tomorrow :-)"}
|
47 |
+
|
48 |
### Data Fields
|
49 |
- dialogue: text of dialogue.
|
50 |
+
- summary: one liner human written summary of the dialogue.
|
|
|
51 |
- id: unique file id of an example.
|
52 |
|
53 |
### Data Splits
|
54 |
+
- train: 14732
|
55 |
+
- val: 818
|
56 |
+
- test: 819
|
|
|
57 |
|
58 |
## Dataset Creation
|
59 |
### Curation Rationale
|
60 |
In paper:
|
61 |
+
In the first approach, we reviewed datasets from the following categories: chatbot dialogues, SMS corpora, IRC/chat data, movie dialogues, tweets, comments data (conversations formed by replies to comments), transcription of meetings, written discussions, phone dialogues and daily communication data. Unfortunately, they all differed in some respect from the conversations that are typically written in messenger apps, e.g. they were too technical (IRC data), too long (comments data, transcription of meetings), lacked context (movie dialogues) or they were more of a spoken type, such as a dialogue between a petrol station assistant and a client buying petrol.
|
62 |
+
As a consequence, we decided to create a chat dialogue dataset by constructing such conversations that would epitomize the style of a messenger app.
|
|
|
63 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
64 |
### Who are the source language producers?
|
65 |
linguists
|
66 |
### Who are the annotators?
|
67 |
language experts
|
68 |
+
### Annotation process
|
69 |
+
In paper:
|
70 |
+
Each dialogue was created by one person. After collecting all of the conversations, we asked language experts to annotate them with summaries, assuming that they should (1) be rather short, (2) extract important pieces of information, (3) include names of interlocutors, (4) be written in the third person. Each dialogue contains only one reference summary.
|
71 |
+
|
72 |
|
73 |
## Licensing Information
|
74 |
non-commercial licence: CC BY-NC-ND 4.0
|